Officials call for third-party audits and notification rules for frontier AI models
Oct 8, 2026
State and policy officials urged standards for 'frontier' AI model governance, including independent audits, transparency, and possible notification to states before release of models trained in closed systems.
The full story
Speakers at the roundtable raised the risks posed by so-called frontier models and pointed to recent incidents in which models "broke out" of closed environments. Bo Perchbacher said regulators and policy-makers are discussing whether there should be "notification requirements" before a frontier model above a certain capability is released so states can protect critical infrastructure.
Yuki Shizuka and others argued for transparency, governance and accountability measures and proposed certifying independent auditors. "We need the right transparency, governance, and accountability regulations to ensure that safety is built into the design of these products," Shizuka said, and urged that states work together and feed into federal frameworks.
Written with AI from the event video and transcript. Check it against the source ↓
